Studies on the Use of Supercritical Ammonia for Ceramic Nitride Synthesis and Fabrication

NASA Technical Memorandum 102570

Linda Cornell and Y.C. Lin Case Western Reserve University Cleveland, Ohio

Warren H. Philipp Lewis Research Center Cleveland, Ohio

April 1990

(_A5 A- l'M-3 025-rt)) 5T_Jr_i r 3 .JU F, -r_,C C, II-ICAL Am_C_NIA r:_n_ C F¢A,'4 IC NIl_ Ir'__ SYNTH=_I$ A_I_; FAjR_ICATI_N CSCt 0 l_ G3125

STUDIES ON THE USE OF SUPERCRITICAL AMMONIA FOR CERAMIC NITRIDE

SYNTHESIS AND FABRICATION

Linda Cornell* and Y.C. Lin* Case Western Reserve Unlversity Cleveland, Ohio 44106

and

Warren H. Phillpp National Aeronautics and Space Administration Lewis Research Center Cleveland, Ohio 44135

ABSTRACT

The extractablIity of ammonium halldes (Including ammonium thiocyanate) formed as byproducts from the synthesis of Si(NH) 2 via ammonolysis of the cor- respondlng sillcon tetrahalldes using supercritical NH 3 as the extraction medium has been investigated. It was found that the NH4SCN byproduct of ammo- noIysls of SI(SCN) 4 can be almost completely extracted with SC ammonia from the insoluble SI(NH) 2 forming a promlslng system for the synthesis of pure Si(NH) 2, one of the best precursors to SI3N 4. In addltion, it was found that Si3N 4, AIN, BN, and Si(NH) 2 are insoluble in SC ammonla. Also discussed are design consideratlons for a supercritical ammonia extraction unit.

INTRODUCTION

Supercritical flulds (fluids above their crltical temperature and pressure) are used as solvents and extraction media because of their near zero surface tension and good solvent properties. Very little has been reported on the use of supercritical (SC) ammonia as a solvent or extractant. This paper pre- sents some prellmlnary results on the use of supercritical ammonia at temperatures and pressures up to 145 °C and 5500 pslg as a solvent for refractory nitride synthesis.

The ability of a supercritlcal (SC) fluid to dissolve low vapor pressure solids was flrst reported by Hannay and Hogarth over lO0 years ago (ref. l). They demonstrated the dlssolutlon of several salts such as cobalt chloride, potassium iodide and potassium chloride in supercritical ethanol (Tc = 234 °C, Pc = 63 atm). In 1896, Villard (ref. 2) published a review of the supercritical solubility phenomena. Carbon dloxide has been extensively investigated as a supercriticaI fluid because of its convenient critlcal temperature, Tc : 31.1 °C, and also because it is nontoxic, nonflammable, environmentally acceptable, and inexpensive. In addltion, it is a good solvent for many organic compounds. SupercritlcaI CO 2 has been used for coffee decaffeination, soybean oii extraction, extractlon of chemotherapeutic agents from plant material and even ethanol-water separation (ref. 3). A general review of supercritical fluid extractions involving a variety of systems is reported in a recent issue of Reviews in Chemical Engineering (ref. 4).

*NASA Resident Research Associate at Lewis Research Center. Supercritical solvents offer four advantages over conventional liquid solvents: (I) enhancedsolubilities, (2) near zero surface tension, (3) high dlffusivity and, thus, high mass transfer, and (4) low viscosity. Enhanced solubilities create a greater driving force for extraction. High dlffusivity, low viscosity, and near zero surface tension of supercritical fluids are advantageous in extraction techniques because the fluid can easily penetrate small pores of a substrate.

Although little information has been reported on supercritical ammonia, supercrltical water has been investigated extensively. For example, the dissolution of sillcon dioxide in SCwater has been demonstrated by Kennedy (ref. 5). The solubility of silicon dioxide in SCwater is enhancedby the presence of a base such as sodium hydroxide. This system is now used to grow single quartz crystals (ref. 6). Surprisingly, very little has been reported on solvent properties of SCammonia.

Basedon an analogy with supercritical water as a mediumin oxide and hydroxide chemistry, supercritical ammoniashould makea good solvent for nitrides and amides. The analogous silicon nitride system would be SC ammonia as the solvent and sodium amide, NaNH2, as the corresponding base. It was this idea that prompted our investigation of the feasibility of using SCammoniafor ceramic nitride processing.

Currently, high purity silicon nitride is synthesized via the preparation of the silicon d11mideprecursor. Thermal decomposition of silicon diimide to s111con nitride occurs according to the reaction:

3Si(NH)2 = Si3N4 + 2NH3 (1) The pyrolysis of silicon diimide to silicon nitride is the basis of the Ub_ process (ref. 7) from which high grade commercial silicon nitride is produced. The precursor silicon diimide is madeby the reaction of silicon tetrachloride with dry ammonia.

SiCI4 + 6NH3 = Si(NH)2 + 4NH4CI (2)

The ammonium chloride is removed by extraction with liquid ammonia, and the insoluble silicon diimide is then thermally decomposed to silicon nitride.

Complete removal of the chloride from the dlimide requires numerous extractions with liquid ammonia or long time perlods of Soxhlet extraction with liquid ammonia (ref. 8). The remaining chloride impurity Interferes with the pyrolysis of the slllcon dlimide because a series of reactions occurs requirlng a higher temperature to remove the chlorlne from the final product, silicon nitride (refs. 9 and I0). In order to synthesize pure silicon nitride at a lowered temperature, it is necessary to devise a more efficient way to extract the ammonium chlorlde byproduct. Extraction wlth supercritical ammonia was investigated with thls idea in mind.

Other systems of silicon dilmide synthesis may lead to purer silicon nitride product. Our preliminary work showed that in addltion to the ammonolysis of silicon tetrachlorlde to produce silicon dlimide, other reactions, equations (3) and (4), also may be used. SIBr 4 + 6NH3 = SI(NH) 2 + NH4Br (3)

Si(SCN) 4 • 6NH3 = SI(NH) 2 + NH4SCN (4)

Although not demonstrated experimentally, it is assumed that the reaction of silicon tetralodide with ammonia also forms silicon diimide and the corresponding ammonium halide, NH41.

In contrast to the reaction of SiCl 4, SiBF 4 and Si(SCN) 4 with ammonia, no sillcon diimide forms in reactions of ammonia with silicon tetrafluoride or silicon tetraacetate. The conventional ammonolysis of silicon tetrafluoride, SiF4, leads to the formation of a volatile solid adduct, SiF4.2NH 3, which on sublimation yields no siiicon diimide or silicon nitride. The addition product of silicon tetraacetate, Si(OCOCH3) ¢ and ammonia forms silicon dioxide instead of silicon nitride on thermal decomposition. Because these two compounds do not lead to silicon nitride, they were deleted from our SC ammonia investigation.

This paper Includes descriptions of the experimental supercritical ammonia apparatus, procedures for the nitride solubility studies, and the studies of byproduct extraction in the synthesis of pure silicon diimide. The results of the nltride solubility studies are discussed. Then the results of SC ammonia extraction of byproduct ammonium salts (NH4CI, NH4Br or NH4SCN) from silicon dlimide are presented along with those from the extraction of NH4CI from porous alumina beads. In addition, equipment design improvements are discussed.

EXPERIMENTAL

Nltride Solubility Experiments

The apparatus used to determine the solubilities of nitrides in supercritical ammonia (SC ammonia) consists of a modification of a high pressure autoclave (an Autoclave Engineers 300 cc vessel). A small crystallizing dish is suspended inside, in the top portion of the chamber well above the liquid ammonla level. The bomb is heated to 145 °C which is above the critical temperature of ammonia, Tc(NH 3) = 132 °C. Nhen a compound such as NH3 is in its supercritical state, it occupies the complete volume of the container. If any nitride dissolves in the SC ammonia, some of the nitride will deposit in the suspended crystallizing dish when the temperature is lowered below the critical point and the liquid ammonia condenses in the bottom of the bomb. A deposit in the crystallizing dish would represent a soluble species. If none dissolves, then all of the nitride would remain at the bottom of the bomb. SEM and optical microscope photographs of the nitrldes were examined before and after exposure to SC ammonia to detect any changes in particle structure which may indicate solubillty and recrystallization of the nitrides.

First, a quantity of the nitride, alpha or beta Si3N 4, AIN or BN Is added to the bomb, the bomb is then closed, and the specified amount of ammonia is condensed on top of the nltride. The solubility of Si3N 4 was also investigated in SC ammonia containing the strong base sodium amide, NaNH2. Thls is accomplished by addlng about 1 g of sodlum shavlngs to the Si3N 4 already in the autoclave. The chamber is immedlately closed to minimize atmospheric reaction wlth the sodium. By the time supercrltlcal conditlons are reached, the sodlum is converted to NaNH2 according to the reaction

3 2Na + 2NH3 = 2NaNH2 + H2 (5) The conditions of pressure and temperature in the bomb, usually 145 °C at 4000 psig, were obtained by condensing the appropriate weight of NH3 into the bomb, then heating the bombto the desired temperature. For the specified conditions of temperature and pressure, the weight of ammoniato be condensed in the bombcan be calculated from the total volume of the container and the density of SCammoniaat that temperature and pressure. At the critical point p(NH3) = 0.235 g/cm3 and at 145 °C and 4000 psig, p(NH3) : about 0.4] g/cm3 (ref. II). The welght of ammoniacondensedin the bombwas determined gravimetrically by difference.

Supercritical AmmoniaExtraction Unit The supercritical ammoniaextraction unit, illustrated in figure I, is a modified version of the Autoclave Engineers supercritical carbon dioxide extraction unit. Ammoniais circulated via a positive displacement liquid metering pump(maximum P = 6000 psig) from the fluid supply cylinder to the system. Instead of the dip tube arrangement shownin figure I, the supply system consisted of a small inverted cylinder from which liquid flows to the pump via its own vapor pressure. Constant pressure is maintained by a back pressure regulator. Becausethe heat of vaporization and the critical temperature of ammoniaare higher than for carbon dioxide, a larger preheater vessel is required to vaporize the liquid ammoniaand heating tapes are added to the system to maintain the specified temperatures above the critical point. The extractor vessel was retrofitted with screens and glass wool to hold the solid sample in place during extraction. Upflow or downflow of solvent through the extractor vessel can be accomplished by appropriate valve adjust- ments. In all of our extractions we used the upflow mode. After extraction, the supercritica] fluid solution flashes across a heated expansion valve into the separator vessel at atmospheric pressure where the dissolved solutes precipitate and the supercritical solvent expands to a gas which Is vented through a flow meter to a fume hood.

Pressure is measuredon both the high and low side of the pump. Tempera- tures are measuredat many points along the system as shownin figure I. Tem- perature is controlled by manually adjusting the appropriate reostat settings and the pressure is also adjusted manually using the back pressure regulator.

Preparation of Samples Silicon diimide was synthesized by ammonolysisof a silicon tetrachloride or silicon tetrabromide solution in heptane, or silicon tetraisothiocyanate in toluene. Becauseof the hydrophilic nature of these silicon compounds,the procedures were carried out in a dry box where possible and care was taken to minimize exposure to atmospheric moisture. The ammonolysisof the sillcon tetrahalide or tetraisothiocyanate yields a 4:1 molar ratio of the ammonium salt of the corresponding anlon to silicon dilmide, which is the precursor to silicon nitrlde. 90 wt %of the NH4SCNwas removedfrom the mixture formed by ammonolyslsof the Si(SCN)4 by extraction with acetonitrile saturated wlth NH3 at 0 °C before SCextraction. This technique was not applicable to the ammonium halides. Pellets of the mixtures were manually pressed in a dry box to facilitate ease of handling. In addition to the salt and diimide, the pellets also contain someadsorbed organic solvent. To determine the extractability of NH4CIwith supercritical ammoniain the extraction system, ammoniumchloride NH4CIwas first absorbed into porous alumina beads. The alumina beads were dried at 144 °C for 24 hr, then vacuum impregnated with 5 M aqueous ammoniumchloride and then dried again overnight at llO °C. The amountof halide absorbed into the pellets was determined gravimetrically. Reagent grade or equivalent chemicals were used. Liquid ammonia,anhy- drous 99.99 percent, was supplied by Liquid Carbonic Corp. The s]licon halides and isothiocyanate were obtained from Petrarch Systems, Inc. Silicon nitride, 325 mesh, 90 percent beta form and 325 mesh, 90 percent alpha form, used in the solubility experiments, was furnished by CERAC.

Extraction Procedures

The supply cylinder is filled with approximately 500 g of ammonia by dis- tillation from a large cylinder rather than by direct liquid transfer. Distil- lation removes nonvolatile solids from the ammonia. To load the extraction vessel, it is removed from the system into the dry box where it is packed with sample pellets and quickly replaced back into the system. Because the silicon dilmide/ammonium halide pellets are sensitive to atmospheric moisture, dry nitrogen was allowed to flow through the extraction section of the system to minimize atmospheric contamination until ammonia flow begins.

Initially, the system is filled with ammonia, then brought to equilibrium at the desired pressure and temperature under no-f_ow conditions. When the system pressures and temperatures are stabilized, the expansion valve is opened to the desired ammonia flow rate and extraction begins. The supercritical fluid solution is allowed to expand into the separation vessel where the solute is collected. The expansion valve and the back pressure regulator are manually adjusted throughout the extraction to maintain the desired temperatures and pressures.

For the silicon diimide/ammonium hallde pellets, the extractions were done in batch wlse steps until no significant amount of additional material was deposited in the separatlon vessel. Thus, durlng each extraction, the system was shutdown two to four times without removing the extraction vessel to evalu- ate the precipitant in the separation vessel. On the other hand, in experiments involving extraction of NH4CI absorbed on the inert porous alumina where quantltatlve rates are required, the extractions were carried out for specified times at a constant supercritical ammonia flow rate.

On completion of each extraction run, the extraction vessel was quickly removed to the drybox for sample recovery. The total amount of precipitant collected in the separator vessel, essentially ammonlum salt, was determined by welghing in a closed weighlng bottle. The composition was verified by x-ray diffraction and the quantity of halide was determined gravimetrically as the silver halide. Trace amounts of silicon were determined colorimetrically using the ammonium molybdate reagent. The sample before and after extraction was evaluated for both halide and s111con content by chemlcal precipitation methods. The sillcon dllmlde was precipltated as silicon dioxide by hydrolysis in d|lute aqueous ammoniafol- lowed by Ignltion in air at 1000 °C. The ammoniumhalide was precipitated as the corresponding silver salt with silver nitrate. The presence of silicon dilmide In the pellets was confirmed by the presence of sillcon nitride after TGAanalysis. Fromthis analysis, the composition of the samples before and after extraction can be determined. Silicon diimide in those extracted pellets derived from SI(SCN)4 was determined by TGA(I0 °C/min in N2 up to 1200 °C) weight loss and the fact that the residue contained Si3N 4 and no other detecta- ble phases were verified by x-ray diffraction.

RESULTS AND DISCUSSION

Nltride Solubillty Experiments

At first SC ammonia was investigated for use in processing ceramic nltrides based on dissolution of nitrldes, especially Si3N 4 in SC ammonia. This idea was based on the analogous supercritlcal water/SiO 2 system whereby solutions of SIO 2 in SC water are used to grow large single crystals of quartz (ref. 5). Unfortunately, no evidence was found that indicated any solubility of SI3N 4, AIN and BN in SC ammonia after 12 hr at 145 °C, 4000 psig. This was shown by the observation that no nitride was found in the crystallizing dish suspended inside of the bomb chamber at the end of the run. When the strong base, NaNH 2 was added to SC ammonia containing SI3N 4, a thin film was seen on the crystallizlng dish. This highly alkaline film was found to contain no silicon and was probably sodium hydroxide derlved from the immediate hydrolysls of the NaNH 2 when the opened bomb was removed from the dry box. Apparently, NaNH 2 is slightly soluble in SC ammonia. The insolubility of Si3N 4, AIN and BN in SC ammonia was verified by examination of SEM microphotographs; no difference in particle morphology was observed before and after exposure to SC ammonia for 12 hr at 145 °C and 4000 psig. An example of the particle structure before and after exposure is shown in figure 2. The lack of solubility is indicated by the retention of the rough edges on the nltrlde particles after exposure to SC ammonia. Even if the nitrides were sllghtly soluble in SC ammonia, the rough edges would be eroded or smoothed out. In cases of moderate solubility, some well defined crystals would be observed due to recrystallization.

In extraction runs presented later, even silicon diimide, Si(NH) 2, was found to be insoluble in SC ammonia. Based on analogy with aqueous chemistry where hydroxides or hydrated oxides tend to be more soluble in water than anhy- drous oxides, one would expect that amides, imides or ammoniated nitrides would be more soluble in SC ammonia than unsolvated nitrides. Because silicon diimide is not soluble in SC ammonia, the use of SC ammonia as an extraction vehl- c]e for the purification of SI(NH) 2 prior to pyrolysis and formation of Si3N 4 was investigated.

Purification of Si(NH) 2 Via SC Ammonia

Si(NH) 2 is prepared by the reaction of the silicon tetrahalides: SiCl 4, SlBF 4, SII 4, and the tetraisothiocyanate, Si(SCN) 4 with ammonia according to the equations SIX4 + 6NH3 _ SI(NH) 2 + 4NH4X (6)

X = CI,Br,I,SCN (7)

The function of the SC ammonia is to extract the ammonium salt byproduct and leave the insoluble SI(NH) 2 residue. We antlcipated high byproduct solubility in SC ammonia because the ammonium salts of interest are very soluble in liquid ammonia.

The extraction results for the Si(NH) 2 + NH4X mixtures are found in table I. The initial sample mixture contains some absorbed organic solvent. For example, the pellets that contain NH4CI also contain about 24 percent toluene. In pellets containing NH4Br, the toulene accounts for about 13 percent of the welght of the pellets, whereas those containing NH4SCN contain 46.5 percent by weight acetonitrile. Thus, the reported composition of each sample is normalized based on the molar percent of Si(NH) 2 and NH4X without solvent.

The table gives the mole ratlo of halide to silicon before and after extraction. Prior to extraction this mole ratio should be X/Si = 4.00. For the halldes SICI 4 and SIBr 4, the ammonolysls was done in toluene solution, but for Si(SCN) 4, acetonitrile, CH3CN, was used as the solvent. Most of the NH4SCN (90 percent by weight) was removed because of its solubility in CH3CN + ammonia. In preliminary experiments, it was found that NH4SCN is soluble in polar organic liquids such as pyridine, tetrahydrofuran and CH3CN all saturated with ammonia. During ammonolysis of SI(SCN) 4 dlssolved in CH3CN most of the NH4SCN dissolved in the CH3CN/NH 3 solution leaving the SI(NH) 2 precipitate with most of the NH4SCN removed. The solubility of NH4SCN in polar organic liquids containing ammonia should be anticipated because of Its high solubility in liquid ammonia, 312 g per lO0 g of ammonia at 25 °C under unspecified pressure (ref. 12). For this case, the initial pseudo halide to silicon ratio used in the SC ammonia extraction was 0.185.

The presence of an oily substance in the separator vessel after the first stage of extraction indicates that SC ammonia removes the organic solvent in addition to the ammonium salt. Also observed in the NH4CI run was a brown dis- coloration of the Si(NH) 2 residue in the extractor. After removing the silicon species as SiO 2 from the acid hydrolysis of the Si(NH) 2 residue, addition of NH40H to the aqueous solution precipitated a rust colored substance identified as hydrated Fe203. The amount corresponds to a Fe/Si mole ratio of 0.236 which is equivalent to about 21 percent by weight. The impurities account for 20 to 30 wt % of the extracted residue. The iron contamination is attributed to attack of the stainless steel sample supports in the extractor by the acid solution of NH4CI in SC ammonia. Rustlng of these screens was also noted.

For the ammonium chloride pellets, the normalized mole percent of Si(NH)2 increased from 19.1 to 78.9 percent and the NH4CI decreased from 80.9 to 21.1 percent after extraction with supercrltical ammonia at 145 °C and 5000 psig. The extraction process removed 93.2 percent of the ammonium chloride. For the ammonium thiocyanate pellets, the normalized mole percent of Si(NH) 2 increased from 84.4 to I00 percent and the NH4SCN decreased from 15.6 to 0 percent resulting in 100 percent of the salt removed. For the ammonium bromide pellets, the normallzed mole percent of Si(NH)2 increased from 19.8 to 69.4 percent and the NH4Br decreased from 80.2 to 30.6 percent resulting In an 89 percent removal of the ammonium bromide. Table II gives data on an estimate of the utilization efficiency of SC ammoniato extract NH4CI, NH4SCN,and NH4Brrespectively. Unfortunately, losses within the system, NH3 loss at the pump, and awkwardconstruction of the apparatus precluded a rigorous quantitative material balance.

The extraction efflciency, especially concerning NH4Cland NH4Br, is low In relatlon to their high solubility in liquid ammonia. The low extraction efficiency for these ammoniumhalides maybe due to the relatively low density of SCammoniaat 145 °C, 5000 psig compared to that of ]iquid ammonia; the solubility of these salts in SC ammonia is not as great as expected. This conclu- sion is also supported by our next set of experiments concerning the extraction of NH4CI absorbed in porous alumina beads.

The removal efficiency for the thiocyanate system is about 150 percent higher than for both the chloride and the bromide containing mixtures. Unlike the chloride or the bromide mixture, we also show that the NH4SCN can be completely removed.

A factor contributing to the relatlvely low solubility of NH4CI and NH4Br in SC ammonia may be the formation of a less soluble triammonlate in the presence of excess ammonia. Morgan (ref. 13), in his investigation of the extraction of NH4CI from the Si(NH) 2 + NH4CI mixture formed by ammonalysis of SiCl 4, found that even a week of continuous Soxhlet extraction with liquid ammonia dld not completely remove all the Cl-ion from the Si(NH)2; O.l percent remained. He suggested that in the presence of ammonia, NH4C] forms a triammoniate which is less soluble in NH 3. We also do not exclude the idea that some of the halide may be chemically bonded to the silicon species making it more difficult to extract.

According to our evaluation of the three routes to Si(NH) 2, the most promising for the preparation of this precursor to Si3N 4 is via the ammonolysis of Si(SCN) 4. Our results show that the pure precursor material can be made by this method Followed by the extraction of the readily soluble NH4SCN.

The bulk of the NH4SCN byproduct was removed by decantation of the NH3-CH3CN solution from the Si(NH) 2 precipltate; then the remainder of the NH4SCN was extracted with SC ammonia. On addition of water to the SI(NH) 2 residue, hydrolysis occurred with the formation of insoluble Si02.XH20. After removal of the Si02.XH20 precipitate, addition of AgNO3 to the acidified (with HNO3) clear supernatent llquid, gave no visual evidence of a AgSCN precipitate thus indicating that no significant of NH4SCN remained in the extracted Si(NH) 2 residue. Based on the handbook value (ref. 14) solubility product of AgSCN, is 1.16xlO -12 at 15 ° C.

Figure 3 gives the TGA of the pure Si(NH) 2 derived from the ammonolysis of Si(SCN) 4 dissolved in CH3CN. The TGA shows a gradual loss in weight up to 1450 °C at which point the sample is 81.4 percent of the original weight. This corresponds to a weight loss of 18.6 percent which is slightly less than the calculated theoretical weight loss of 19.5 percent when Si(NH) 2 is thermally converted to Si3N 4 according to the following reaction.

3Si(NH) 2 = SI3N 4 + 2NH3 (8)

8 The dashed TGAcurve on figure 3 is the TGAobtained by Morgan (ref. 13) for SI(NH)2 obtained from SiCI4 and NH3. The NH4CIcontent was less than O.l percent. Both curves showa gradual weight loss with no indicatlon of well defined intermedlates. Other differences to be noted are that Morgan's Si(NH)2 was recovered from liquid ammoniawhereas ours was exposed to a higher temperature (145 °C supercrltical NH3). The total weight loss, 23 percent, of Morgan's SI(NH)2 exceeded the theoretical 19.5 percent for the decomposition of pure SI(NH)2 to SI3N4. This may be due to someNH3 of solvation contalned in Morgan's material which would add to the weight loss whereas our material, whlch was exposed to SCammoniaat a higher temperature of 145 °C intermediates forming nitride-imides of the type Si2N2NH may have lost some of its NH 3 due to decomposition. This would give a less than stoichiometric weight loss.

On comparison with Morgans TGA which shows that Si(NH) 2 is completely decomposed to the nltride above llO0 °C, we can assume that, in our TGA, all the diimlde is converted to nitride at 1450 °C. Between this temperature and 1500 °C, the end of the TGA run, there is a sudden additional weight loss of 1.6 percent resulting in a final value of 79.8, corresponding to a 20.2 percent weight loss. However, in a separate experiment no additional weight loss occurred upon heating of the Si(NH) 2 from 1500 to 1700 °C. This suggests that the TGA curve (fig. 3) would probably flatten out above 1500 °C. Although we are unable at this time to establish a definite cause for this sudden weight drop, this result Is not thought to be characteristic of Si(NH) 2 decomposition. Rather, we attribute this loss to the vaporization of SiO as a consequence of SiO 2 impurity in the sample; Si(NH) 2 and its decomposition product intermediates are extremely sensitive to hydrolysis; thus, any reaction with atmospheric moisture during handling would result in SiO 2 contamination in the final product.

The residue remaining after the TGA was amorphous by x-ray diffraction. However, when the sample was heated to 1700 °C for 8 hr in N2, the amorphous material was converted to pure alpha Si3N 4 according to x-ray diffraction analysis. Based on this result, it is logical to assume that the end-product of the TGA run was amorphous Si3N 4.

The synthesis of Si(NH) 2 from Si(SCN) 4 and NH3 has advantages for the preparation of S_3N4 at a relatively low temperature which favors the formation of finely divided powder with high surface energy. Such powders are easily densified into useful artlcles.

Extraction of NH4Cl From Porous Alumina Beads

To understand the effect of solvent parameters on extraction efficiency, a simple system involving NH4CI absorbed on to the inert substrate, porous alumina, was investigated at four SC ammonia conditions and one liquid ammonia parameter. These results are presented in table Ill and illustrated by plot- ting extraction efficiency, the percent NH4CI extracted, versus solvent density in figure 4. On examination of figure 4, it becomes apparent that solvent density influences extraction efficiency. The greatest extractlon efficiency (91.6 percent NH4CI removed) was obtained at the highest density conditions tested, i.e., in liquid NH 3 at 50 °C and 4000 psig. The lowest efficiency (52 percent NH4CI removed) occurred at the lowest density conditions, near the supercritical point of NH 3 at 145 :C and 2000 psig. Increasing the density of the extractant, NH3, enhances the extraction efficiency by increasing the solubility of the solute in the solvent.

There is considerable discussion explaining nonionic solute solubility in essentially nonpolar to slightly polar supercritical fluids, e.g., naphthalene in ethylene, and benzoic acid in CO2. McHugh and Krukonic (ref. 15) describe a relationship based on solute sublimation pressure to predict the solubillty of nonvolatile, covalent solids In high density SC nonpolar fluids.

Unfortunately no such empirical relationshlp has been well developed to predict solubility of ionic crystalline solutes e.g., NH4CI in polar supercrit- Ical solvents such as ammonia.

In order to dissolve an ionic crystal, NH4Cl, in a polar solvent, NH3, there must be an interaction between solute and solvent, which is solvation of the anion and cation, to break up the crystal into individual ions. Thus, the higher the solvent concentration, the more solvent molecules are available for interaction. With this fact in mind, the solubility of ionic solutes is expected to increase with increase in SC solvent density.

However, density or solubility is not the only parameter affecting the extraction efficiency as evidenced by its minimal change with density between points B and D. Mass transport properties such as dlffusivity and viscosity also influence the system behavior. For example, although the density at point D is greater than at point C, point D extractlon efficlency is slightly less. But the dlffusivity is greater and the viscosity Is lower for point C than for point D. Thus, the more favorable mass transport properties at condi- tion C favor extraction when compared with conditions at point D. In the supercritical region between points B and D it appears that the magnitude of the mass transport effects are the same order or greater than the solubility effects due to density whereas at the two endpoints, A and E, the density of the extraction media, NH 3, dominates the extraction process.

It is evident from our data that extraction in the high temperature liquid phase is more efficient in terms of solute extracted than at any of the supercrltical conditions tested. However, supercritical conditions at much higher pressures may produce better extraction efficiency than in the liquid state as the supercritical fluid approaches near liquid density while maintaining more favorable transport properties.

Equipment Design Improvements

Because of the absence of reported prior art on the design of a supercritical ammonla extractor unit, several shortcomings of our equipment design be- came evident through use experience. These problems and suggested improvements are discussed below. These design improvements may be useful to others inter- ested in building a supercritical ammonia extractor for future investigations.

One troublesome problem was the frequent development of ammonia leaks in the pump caused by small scratches in the sapphire plunger of the pump. The scoring of the plunger was caused by small carbon particles in the NH 3 stream released when the carbon-fllled teflon O-ring degraded. The leaks were pre- vented by replacing the old seals with polypropylene O-rings without filler.

I0 Wefound that the poIypropylene O-rings were considerably more resistant to attack by ammoniathan those comprised of teflon with a carbon filler. Because of the high pressure dlfference between the high and low sides of the pump, even minute scores in the pumpplunger cause significant ammonialeaks. It therefore becomesimperative that the ammoniafeed be free of particu]ates; u]tra-pure ammoniafeed stock must be used, an adequate filter to remove the fine particles must be provided, and all O-ring seals should contain no solid filler material.

Cavltation also occurred in the pumpbecause the intake pressure was not high enough to continuously feed liquid ammoniato the pump. Weused a heat gun as a stop gap measureto warm the ammonlafeed reservoir to increase the intake pressure. A controlled heater should be added to the system to heat the reservoir. In addition, the intake pressure and temperature should be monitored.

The orlglnal design, based on a rm_dified supercritical CO 2 extractor, provided a larger than necessary extractor vessel volume. This was done because most SC fluid applications involve extraction of a small amount of matter from a large volume of material, as is the case with the removal of caffeine from coffee beans. Our requirements are the opposite; we are extracting a large proportion of soluble compound from a comparatively limited volume of material. For our use, a smaller extractor vessel volume would be advantageous. Also, the separator should be located immediately after the extraction vessel, the tubing lengths could be reduced and some valves eliminated by providlng upflow only through the extractor. These changes would a11ow for a more compact system with shorter connecting tubing lengths between extraction and separator vessels and fewer valves thereby minimizing clogging and material losses due to deposition of solute in the tubing. To prevent plugging of the expansion valve due to extremely soluble species, the solutions could be diluted with pure ammonia which is at the solution temperature and pressure at the outlet of the extractor vessel.

Automated temperature, pressure, and flow feedback control systems would aid in maintaining steady state condltlons. In addition, the heating tapes currently in use should be replaced by placing the system in an automatically controlled oven to provide a constant heat source. This would maintain a more constant temperature over the system.

Finally, solutions of ammonium halides in ammonia are corrosive because they behave as strong acids. Corrosion of the stainless steel screens used to hold the sample in the extractor was evident. He suggest that the screens be made of a corrosion resistant material.

CONCLUSIONS

The solubility of ammonium chloride in SC ammonia appears to be largely a function of solvent density as shown by the increased extractlon efficiency at higher SC ammonia pressure rather than being temperature dependent. In fact, maximum extraction capability was obtained using high temperature (50 °C) liquid ammonia. The higher the solvent density the greater the solvent activity in terms of solvent molecules per unit volume. Thus, a high solvent density allows greater interaction between the ionic solute and solvent molecules,

II causing increased solubility. Mecould assumethat the solubillty of most Inorganic compounds would go up as the SC ammonia pressure is increased. Unfortunately, our bomb and extraction systems were pressure llmited to under 5000 psig. It is posslble that at conslderably higher SC ammonia pressure, nltrldes, and especially inorganic amides and imides would be more soluble. In our studies, $i3N 4, AIN, BN, and Si(NH) 2 were Found to be insoluble in SC ammonla. At such high SC ammonia pressures, reactions of nltrides with the strong base NaNH 2 to form soluble complexes may also take place much in the same man- ner as the formation of soluble species from $i02 and NaOH occurs in supercrit- Ical water.

Concerning the extraction of the ammonium salt formed as a reaction product from the ammonolysls of the corresponding silicon species, SC ammonia at a higher pressure would probably afford a better separation than we found at 4000 pslg. The only highly effective separation that we report with SC ammonla was the extraction of NH4SCN from Si(NH) 2 formed by reaction of Si(SCN) 4 with ammonia. Probably this could have been accomplished just as well with llquld ammonia. We belleve that Si(NH) 2 is the best precursor for the production of high purity, finely dlvided, easily sinterable SI3N 4. Based on our results one of the best routes to the synthesis of pure SI(NH) 2 is via the ammonolysis of Si(SCN) 4 dissolved in an organic solvent. Compared to other sillcon starting materials, namely the tetrahalides, the byproduct, NH4SCN is more efficiently removed by SC ammonla and probably liquid ammonia extractlon than the reaction byproducts from the ammono]ysis of SlC] 4 and SiBr 4, which are NH4Cl and NH4Br, respectlve]y.

REFERENCES

I. Hannay, J.B.; and Hogarth, J.: On the Solubility of Solids in Gases. Proc. R. Soc. London, vol. 29, 1879, pp. 324.

2. Vlllard, P.: Solubility of Llquids and Sollds in Gas. J. Phys., vol. 5, 1896, pp. 455.

3. McHugh, M.A.: and Krukonis, V.J.: Supercritical Fluid Extraction: Principles and Practice. Butterworth Publishers, Stoneham, MA, 1986, Chapter lO.

4. Paulaltis, M.E.; et al.: Supercritical Fluid Extraction. Rev. Chem. Eng., vol. I, no. 2, Apr-June 1983, pp. 179-250.

5. Kennedy, G.C.: A Portion of the System Silica-Water. Econ. Geol., vol. 45, no. 3, 1950, pp. 629-653.

6. Laudise, R.A.: Hydrothermal Synthesis of Crystals. Chem. Eng. News, vol. 65, no. 39, Sept. 28, 1987, pp. 30-35, 38-43.

7. lwal, T.: Kawahlto, T.; and Yamada, T.: Process for Producing Metallic Nitride Powder. U.S. Patent 4,196,178, Apr. I, 1980.

8. Morgan, P.E.D.: Research on Denslflcatlon, Character and Properties of Dense Sillcon Nltrlde. FIRL-A-C3316, The Franklin Institute Research Laboratories, Philadelphia, PA, 1973 (Avall. NTIS, AD-757748).

12 9. Bllly, M." Preparation and Definition of Silicon N1tride. Ann. Chim. (Paris), vol. 4, no. 13, 1959, pp. 795-851. 10. 8111y, M., et a1.: Synthesis of Si and Ge Nitrides and $i Oxynitride by Ammonolysisof Chlorides - Commenton "Synthesis, Characterization, and Consolidation of $13N4 Obtained from Ammonolysisof SIC14." J. Am. Ceram. Soc., vo1. 58, no. 5-6, May-June 1975, pp. 254-255. 11. Pressure Enthalpy Diagram for Refrigerant 717. ASHRAEHandbook, 1985 Fundamentals, SI Edition, ASHRAE,1985, p. 17.32, fig. 15. 12. Sneed, M.C.; Maynard, J.L.; and Brasted, R.C." SomeSolubilities of Inorganic Salts In Ammonia. ComprehensiveInorganic Chemistry. Vol. 5, Van Nostrand, 1956, p. 160, Table 3.2. 13. Morgan, P.E.D." Production and Formation of Si3N4 From Precursor Materials. FIRL-A-C3316-5, The Franklin Institute Research Laboratories, Philadelphia, PA, 1973, p. 8. (Avail. NTIS, AD-778373). 14. Handbookof Chemistry and Physics, 1982-1983, 63rd edition, R.C. Weast, ed., CRCPress, Inc., Boca Raton, FL, 1983, 8-242.

15. McHugh,M.A.; and Krukonis, V.J." Supercritical Flu_d Extraction - Principles and Practice. Butterworths Publishers, Stoneham, MA, 1986, Chapter 5, p. 89.

TABLE I. - COMPOSITIONS AND BYPRODUCT REMOVAL IN Si(NH) 2 SYSTEMS

System Initial composition Final composition Halides removed, Normalized, Molar ratio, Normal ized, Molar ratio, percent mol % X/Si mol % XlSi

Si(NH)2+NH4C1 Si(NH) 2 19.1 4 78.9 0.246 93.2 NH4CI 80.9 21.1

Si(NH)2+NH4SCN Si(NH) 2 84.4 .185 100 0 100 NH4SCN 15.6 0

Si(NH)2+NH4Br Si(NH) 2 19.8 69.4 .44 89 NH48r 80.2 30.6

TABLE II. - AMMONIA USAGE

System Steady state Removal efficiency, percent

Time, NH 3 , NH4 salt (g) Percent halide hr g removal per gram NH3 (g) of HN 3

Si(NH)2+NH4CI 6 600 3.856xi0 -3 0.155

Si(HN)2+NH4SCN 4-I/3 250 2.82 .40

Si(NH)2+NH4Br 6 555 16.69 .16

13 TABLE III.- SUPERCRITICAL AMMONIA EXTRACTION [NH4CL in Al203 pellets.]

Run number a

E-I E-2 E-4 E-5

Experimental conditions Total NH 3 flow, ACF 17.05 16.38 17.06 17.04 15.32 Pressure, psi 4000 2000 5500 4000 4000 Temperature, °C 145 145 145 50 180

Weight of pellets, g Before extraction 17.5882 17.2687 17.4014 17.4378 17.2267 After extraction 15.9753 16.2922 15.8643 15.8975 16.2555

Weight of NH4C1, g Extracted 1.629 0.9765 1.5371 1.5403 0.9712 Unextracted 0.7279 0.9002 0.7430 0.1412 0.5487 Total 2.3408 1.8767 2.2801 1.6815 1.5199

NH4CI extracted, percent 68.9 52.0 67.4 91.6 63.9

NH4CI in separator, g 1.2069 0.0072 0.7184 1.9149 0.0186

NH4Cl/gram of pellets 0.1331 0.I087 0.1310 0.0964 0.0882

Overall ratio (estimated) 0.1377 0.1408 0.1408 0.1408 0.1408

aExcept run E-I pellets used are from the same batch. bln a period of 2-I/2 hr.

PRESSURE PRESSURE GAUGE GAUGE I r _) TEMPERATUREREAD( )UT i II CHECK PRESSURE W_ I ) RUPTURE DISC GAUGE I i I _ TEMPERATURE T .,_ RELIEF HEAT I__ALVE SOURCE L--J _11

FLUID SUPPLY METE R CYLINDER SEPARATOR VESSEL

TOTALIZERFLOW LIQUID PHASE SAMPLE TO EXTRACTOR_--VESSEL .d , _ VENT CRITICAL PHASE Figure 1. - Supercdtical ammoniaextractionunit.

14 ORIGINALPAGE BLACK AND WHITE PHO.T_OGRAP_H

(a) Before. (b) Atter.

Figure 2. - AIN powder before and after exposure to supercdtical ammonia.

4000 psig 95 m 5OoC 100 D E REF. 13) 95 m

80 m

-- W I"-- oz 75 --LU 4000 psig 145 °C 5500 psig _30 70 4000 psig 180 °C [] 145 °C c [] [] D - B 80 _ 60 2000 psig 55 -- 145 °C I I I _F---T-- I ADI I I I I I I I 75 50 3O 280 530 780 1030 1280 1530 .260 .300 .340 .380 .4_ .460 .20 ._0 .580 TEMPERATURE, °C DENSITY OF AMMONIA, gm/cc

Figure 3. - Si(NH) 2 from Si(SCN) 4. Figure 4. - Supercritical ammonia extraction; ammonium chloride in alumina beads.

15 National Aeronautics and Report Documentation Page Space Admlnistral_on

1. Report No. 2. Government Accession No. 3. Recipient's Catalog No. NASA TM- 102570

4. Title and Subtitle 5. Report Date

Studies on the Use of Supercritical Ammonia for Ceramic Nitride April 1990 Synthesis and Fabrication 6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

Linda Cornell, Y.C. Lin, and Warren H. Philipp E-4441

10. Work Unit No.

510-O1-0A 9. Performing Organization Name and Address 11. Contract or Grant No. National Aeronautics and Space Administration Lewis Research Center Cleveland, Ohio 44135-3191 13. Type of Report and Period Covered

'12. Sponsoring Agency Name and Address Technical Memorandum

National Aeronautics and Space Administration 14. Sponsoring Agency Code Washington, D.C. 20546-0001

15. Supplementary Notes

Linda Cornell and Y.C. Lin, Case Western Reserve University, Cleveland, Ohio 44106 and NASA Resident Research Associate at Lewis Research Center.

16. Abstract

The extractability of ammonium halides (including ammonium thiocyanate) formed as byproducts from the synthesis of Si(NH) 2 via ammonolysis of the corresponding silicon tetrahalides using supercritical NH3 as the extraction medium has been investigated. It was found that the NHaSCN byproduct of ammonolysis of Si(SCN)4 can be almost completely extracted with SC ammonia from the insoluble Si(NH)2 forming a promising system for the synthesis of pure Si(NH)2, one of the best precursors to Si3N 4. In addition, it was found that Si3N 4, AIN, BN, and Si(NH)2 are insoluble in SC ammonia. Also discussed are design considerations for a supercritical ammonia extraction unit.

17. Key Words (Suggested by Author(s)) 18. Distribution Statement

Supercritical fluids Unclassified - Unlimited Supercritical ammonia Subject Category 25 Silicon nitrides

19. Security Classif. (of this report) 20. Security Classif. (of this page) 21. No. of pages 22. Price* Unclassified Unclassified 16 A03

NASAFORM1626OCT86 *For sate by the National Technical Information Service, Springfield, Virginia 22161

National Aeronautics and FOURTH CLASS MAIL Space Administration illIil

Lewis Research Center ADDRESS CORRECTION REQUESTED Cleveland, Ohio 44135

Official BuslmNm

Penalty for Private Use $300 PoslaQe and Fee_ Paid

Space Adrr,nt,qrah, m

NASA 451